arXiv subjects

Zhichao Yang

Publications and source records attributed to Zhichao Yang.

At least 19 recordsLinked to original sources

arXiv2026

A Hyperbolicity Atlas of Large Language Model Hidden States

LLM hidden states are ordinary vectors, but the distances among those vectors may still show hierarchical structure. To our knowledge, this paper is the first systematic study of whether prompt-token hidden states in contemporary LLMs exhibit Gromov Hyperbolicity (GH), a distance-based measure of tree-likeness. Using 818,904 sample-layer measurements from ten open-weight models across MATH500, HumanEval, WinoGrande, and TruthfulQA, we build a GH map over four axes: parameter scale, layer depth, model family, and input domain. The clearest pattern is depth, not scale: middle layers usually form a high-relative-hyperbolicity plateau, while final layers often become substantially more tree-like. Scale effects are weak and non-monotonic, matched 7/8B model families differ strongly, and domains interact with model specialization. These findings make GH useful as a practical diagnostic: it shows where hierarchical distance structure appears, how specialization changes it, and which model-layer-domain comparisons deserve closer analysis.

cs.CL↗

arXiv2026

When Calibration Rankings Reverse: Accuracy-Controlled Evaluation for Fair Comparison of LLMs

Calibration evaluates whether a model confidence aligns with its empirical accuracy. Existing studies often compare the calibration of different large language models using global calibration metrics such as Expected Calibration Error and Brier Score. We begin by showing, both theoretically and empirically, that such comparisons are confounded by differences in model accuracy. For fairer cross-model comparison, we then propose ACE, an accuracy-controlled evaluation framework with three complementary views: Instance-Aligned, Distribution-Aligned, and Candidate-Aligned calibration. Across multiple benchmarks, model families, and confidence elicitation methods, we use ACE to study two practically important comparison axes, small versus large models and thinking versus non-thinking models. We find that many previously reported calibration advantages under raw global metrics weaken substantially after accuracy control. We also find that ranking reversal is frequent: models favored by raw metrics often cease to be favored once accuracy is controlled. Our results show that raw global calibration metrics are not robust for cross-model comparison, and that fair calibration comparison requires accuracy-aware evaluation.

cs.CL↗

arXiv2026

MedFabric: Gold Evidence Hides the Difficulty of Word-Level Medical Fabrication Detection

Large language models fabricate in medicine, producing fluent statements that are factually wrong, so reliable fabrication detection is a prerequisite for clinical deployment. Reported progress on this task is inflated by two evaluation artifacts: an authorship-style shortcut, where human-written ground truths are paired with LLM-written hallucinations so detectors key on writing style rather than facts, and the provision of gold evidence at test time. A benchmark that tests factual reasoning must therefore remove the style shortcut, ground every fabrication in a real retrievable passage, and score detectors across the range of evidence quality faced in deployment. We build MedFabric to these requirements, a benchmark of 646 word-level medical fabrications, each paired with a ground truth that shares its LLM authorship and near-identical surface form (median ROUGE-L 0.95). On MedFabric the task is unsolved: expert clinicians reach only 53.3% macro F1 and no detector family clears about 60% without gold evidence. Our central finding is that detection is governed by evidence correctness rather than fabrication subtlety, since a strong LLM scores 91% with the gold passage but falls to 35%, below its own no-evidence baseline, under a wrong one, a pattern that holds on two benchmarks at two model scales. The failure is actionable: a retrieval-confidence gate that abstains on low-confidence evidence raises macro F1 from 61% to 74%, and we release MedFabric, all code, and every baseline.

cs.CL↗

arXiv2026

AGIDefect-4K: A Richly Annotated Dataset for AI-Generated Image Defect Detection, Localization and Explanation

Generative AI can now produce highly realistic images, yet current models still exhibit subtle but critical defects that undermine their reliability. While existing AI-generated image (AGI) evaluation benchmarks have made notable progress, comprehensive AGI defect diagnosis remains underexplored. To bridge this gap, we introduce AGIDefect-4K, a richly annotated dataset of 4,000 images from 15 state-of-the-art generative models spanning both open-source and closed-source systems. AGIDefect-4K features hierarchical defect annotations: (1) detection labels identifying whether defects exist, (2) pixel-level segmentation masks localizing defective regions, and (3) detailed textual explanations characterizing defect types and their perceptual impact. Each image is further annotated with an overall quality score. Building on this, we present AGIDA (AGI Defect Assistant), a baseline framework leveraging Multimodal Large Language Models (MLLMs) for joint defect detection, localization, explanation, and quality prediction. Comprehensive benchmarking on AGIDefect-4K reveals that AGI defect understanding remains challenging, underscoring the value of this dataset. The dataset is publicly available at https://github.com/sxfly99/AGIDefect-4K.

cs.CV↗

arXiv2026

Speculative Rollback Correction for Quality-Diverse Web Agent Imitation

Training interactive web agents through imitation learning from expert trajectories has emerged as a highly effective approach. However, determining the optimal timing for expert intervention presents a critical challenge in this context. Delayed intervention often leads to the accumulation of early-stage errors, pushing the page state into an irrecoverable regime. Conversely, premature or excessive intervention causes the agent to become overly reliant on expert policies, trapping the model in local optima characterized by a single, rigid trajectory. We propose Speculative Rollback Correction (SRC), a branch-level imitation framework for resettable agent environments. Instead of requesting teacher labels at every visited state or correcting only after a completed trajectory, SRC uses fixed-horizon branch review: the student executes a short speculative segment before teacher review, and the teacher localizes the first harmful deviation only when local progress breaks. Rollback preserves useful prefixes, while successful rollouts are filtered by a hard verifier and retained in a lightweight quality-diversity archive. The resulting data supports next-action supervised fine-tuning on both localized corrections and verifier-passing trajectories. On WebArena-Infinity, SRC collects 977 verifier-passing trajectories and 9,183 next-action examples; fixed-horizon review improves the recovery-versus-query tradeoff over step-level review while retaining verifier-passing solution variants. Code is available at https://github.com/LongkunHao/SRC_gui_agent.

cs.LG↗

arXiv2026

Enhancing LLMs for Identifying and Prioritizing Important Medical Jargons from Electronic Health Record Notes Utilizing Data Augmentation: A Comparative Study

OpenNotes gives patients access to their EHR notes, but dense medical jargon limits comprehension. We evaluate closed-source and open-source LLMs for extracting and prioritizing the jargon terms most relevant to individual patients, using 90 expert-annotated EHR notes. We test combinations of general vs. structured prompts, zero-shot vs. few-shot prompting, fine-tuning, and GPT-4o-based data augmentation, the last paired with a ranking technique to refine training in low-resource settings. To assess the effect of dataset size, we fine-tune on augmented datasets scaled from 10 to 9,995 examples. All settings are evaluated with 10-fold cross-validation, reporting F1 and Mean Reciprocal Rank (MRR) under two string-matching criteria (relaxed matching and Jaccard Index), followed by an error analysis of model outputs. Open-source models performed best when fine-tuned on the gold-standard dataset: under Jaccard-based matching, DeepSeek 8B achieved the top F1 (0.431, SD 0.046) and BioMistral 7B the top MRR (0.577, SD 0.109). Under relaxed matching, however, open-source models did not match closed-source performance even with augmentation or fine-tuning. Few-shot prompting offered no advantage over zero-shot in vanilla models; prompting style substantially affected results; fine-tuning on a small gold-standard set improved performance; and data augmentation matched or exceeded fine-tuning, though its benefit depended heavily on augmented-data quality. These findings show that prompting strategy, fine-tuning, and data augmentation each meaningfully improve LLM performance on patient-centered jargon extraction in low-resource clinical settings.

cs.CL↗

arXiv2026

Personalized Image Aesthetic Assessment via Preference-rich Sample Mining and Cohort Merging

Personalized Image Aesthetic Assessment (PIAA) aims to predict aesthetic ratings of images that vary across individuals. The aesthetic preferences manifest to different extents across distinct visual stimuli and exhibit cohort-specific patterns. Motivated by the above fact, this paper presents a Multimodal Large Language Model (MLLM)-based approach, which models individual aesthetic preferences by Preference-Rich sample mining and Aesthetically-resonant Cohort merging (PRAC). Specifically, PRAC first identifies preference-rich samples by analyzing both Collective Controversy and Personalized Deviation of images, maximizing the utility of limited user data. Based upon the preference-rich samples, cross-user preference similarities are measured by comparing preference embeddings. Then, a cohort-based model merging strategy, is proposed by aggregating preference patterns from aesthetically-resonant users, which further enhances the personalization for the target individual. Extensive experiments and comparisons on four benchmark PIAA databases demonstrate the superiority of the proposed PRAC model over the state-of-the-arts. The code and model will be public at https://github.com/yzc-ippl/PRAC.

cs.CV↗

arXiv2026

StainFlow: Entity-Stain Tracking and Evidence Linking for Process Rewards in GUI Agents

Reinforcement Learning (RL) has become a promising approach for improving GUI Agents in long-horizon, stochastic digital environments, but trajectory-level success feedback is too sparse to provide reliable credit assignment for intermediate exploration steps. To mitigate this issue, recent studies introduce Process Reward Models (PRMs), which provide finer-grained training feedback through global milestone verification or local step-level evaluation. However, these methods still suffer from two level-specific limitations: global milestone decomposition is subjective and singular, making it difficult to accommodate the multiple valid execution paths in real GUI tasks, while fixed local judging windows may miss long-range key evidence or dilute the decision signal with irrelevant frames. Inspired by stain-tracing mechanisms in network flow analysis, we propose StainFlow, an entity-stain-flow process reward model for GUI Agents. To reduce the subjectivity of global partitioning, we introduce the Global Entity Stain Tracking module, which extracts visually verifiable task entities and tracks how their stain concentrations and states evolve along the trajectory, allowing task phases to be objectively separated by changes in the entity evidence flow. To improve the accuracy of local verification, we introduce the Local Stain Evidence Linking module. Centered on the triggering entities of each candidate key node, it retrieves relevant steps based on their stain concentrations and state changes, and dynamically constructs high-density evidence windows for verifying true key nodes. Extensive experiments on AndroidWorld and OGRBench show that StainFlow relatively improves online RL success by 3.2% and trajectory completion judgment accuracy by 1.8%.

cs.AI↗

arXiv2026

PACT: Learning Diverse Diagnostic Strategies via Privileged Synthesis and Branch Consensus

Clinical diagnosis requires flexible use of multiple reasoning paradigms under incomplete patient information. Existing LLM-based medical agents show strong medical reasoning ability, but single-paradigm or naively mixed dialogue supervision makes these paradigms difficult to learn without interference. We propose \textbf{PACT} (Periodic Anchor Consensus Training), a framework that couples supervised multi-paradigm dialogue synthesis with consensus-based Branch training. At the data level, \textbf{DPS} (Doctor-Patient-Supervisor) uses complete electronic medical records (EMRs) for quality control while keeping the doctor agent restricted to patient-visible information. This produces validated dialogues under four diagnostic reasoning paradigms without leaking hidden clinical answers. At the training level, PACT trains one paradigm-specific LoRA Branch per paradigm and periodically aggregates Branches into a shared Anchor through sign consensus. We further construct a dynamic multi-turn Chinese medical diagnosis benchmark for interactive consultation. Experiments show that PACT achieves state-of-the-art performance among compared proprietary, medical-specialized, and task-adapted baselines on diagnostic outcome and consultation-process metrics.

cs.CL↗

arXiv2026

MIRAGE: Mobile Agents with Implicit Reasoning and Generative World Models

Mobile agents are increasingly expected to operate everyday applications from screenshots and language goals, where reliable control requires reasoning over screen affordances, multi-step navigation, and future state changes. However, many agents externalize this computation as long textual chains of thought, which slows interaction, increases supervision cost, and complicates deployment. We introduce MIRAGE, a framework that learns continuous latent reasoning representations from visible textual reasoning traces. MIRAGE transfers explicit reasoning into compact hidden states, enabling the agent to reason internally without decoding long rationales. It also incorporates a generative world-model objective: latent reasoning vectors are aligned with future screenshots, encouraging the agent to anticipate upcoming interface states before acting. This turns hidden computation into both a compressed thought representation and a forward-looking model of environment dynamics. At inference time, MIRAGE reasons in continuous latent space, reducing token generation while improving execution efficiency. On AndroidWorld, MIRAGE matches explicit chain-of-thought supervised fine-tuning in the 4B ablation with a 3-5x lower decoded-token budget and improves a comparable instruction-tuned baseline by 10.2 points; on AndroidControl, it improves action grounding while generating over 75% fewer tokens.

cs.AI↗

arXiv2026

Medical thinking with multiple images

Large language models perform well on many medical QA benchmarks, but real clinical reasoning often requires integrating evidence across multiple images rather than interpreting a single view. We introduce MedThinkVQA, an expert-annotated benchmark for thinking with multiple images, where models must interpret each image, combine cross-view evidence, and answer diagnostic questions with intermediate supervision and step-level evaluation. The dataset contains 8,067 cases, including 720 test cases, with an average of 6.62 images per case, substantially denser than prior work, whose expert-level benchmarks use at most 1.43 images per case. On the test set, the best closed-source models, Claude-4.6-Opus, Gemini-3-Pro, and GPT-5.2-xhigh, reach only 57.2%, 55.3%, and 54.9% accuracy, while GPT-5-mini and GPT-5-nano reach 39.7% and 30.8%. Strong open-source models lag behind, led by Qwen3.5-397B-A17B at 52.2% and Qwen3.5-27B at 50.6%. Further analysis identifies grounded multi-image reasoning as the main bottleneck: models often fail to extract, align, and compose evidence across views before higher-level inference can help. Providing expert single-image cues and cross-image summaries improves performance, whereas replacing them with self-generated intermediates reduces accuracy. Step-level analysis shows that over 70% of errors arise from image reading and cross-view integration. Scaling results further show that additional inference-time computation helps only when visual grounding is already reliable; when early evidence extraction is weak, longer reasoning yields limited or unstable gains and can amplify misread cues. These results suggest that the key challenge is not reasoning length alone, but reliable mechanisms for grounding, aligning, and composing distributed evidence across real-world multimodal clinical inputs.

cs.CV↗

arXiv2026

State Beyond Appearance: Diagnosing and Improving State Consistency in Dial-Based Measurement Reading

Multimodal large language models (MLLMs) have achieved impressive progress on general multimodal tasks, yet they remain brittle on dial-based measurement reading. In this paper, we study this problem through controlled benchmarks and feature-space probing, and show that current MLLMs not only achieve unsatisfactory accuracy on dial-based readout, but also suffer sharp performance drops under viewpoint and illumination changes even when the underlying dial state remains fixed. Our probing analysis further reveals that same-state samples under appearance variation are not consistently clustered, while neighboring states fail to preserve the local structure implied by continuous dial values. These findings suggest that existing MLLMs largely ignore the intrinsic state geometry of dial measurement tasks and instead rely on superficial appearance cues. Motivated by this diagnosis, we propose TriSCA, a tri-level state-consistent alignment framework for dial-based measurement reading. Specifically, TriSCA consists of state-distance-aware representation alignment, metadata-grounded observation-to-state supervision, and state-aware objective alignment. Extensive ablation studies and evaluation experiments on controlled clock and gauge benchmarks, together with evaluation on an external real-world benchmark, demonstrate the effectiveness of our method.

cs.CV↗

arXiv2026

RiTeK: A Dataset for Large Language Models Complex Reasoning over Textual Knowledge Graphs in Medicine

Answering complex real-world questions in the medical domain often requires accurate retrieval from medical Textual Knowledge Graphs (medical TKGs), as the relational path information from TKGs could enhance the inference ability of Large Language Models (LLMs). However, the main bottlenecks lie in the scarcity of existing medical TKGs, the limited expressiveness of their topological structures, and the lack of comprehensive evaluations of current retrievers for medical TKGs. To address these challenges, we first develop a Dataset1 for LLMs Complex Reasoning over medical Textual Knowledge Graphs (RiTeK), covering a broad range of topological structures. Specifically, we synthesize realistic user queries integrating diverse topological structures, relational information, and complex textual descriptions. We conduct a rigorous medical expert evaluation process to assess and validate the quality of our synthesized queries. RiTeK also serves as a comprehensive benchmark dataset for evaluating the capabilities of retrieval systems built upon LLMs. By assessing 11 representative retrievers on this benchmark, we observe that existing methods struggle to perform well, revealing notable limitations in current LLM-driven retrieval approaches. These findings highlight the pressing need for more effective retrieval systems tailored for semi-structured data in the medical domain.

cs.CL↗

arXiv2026

MedicalBench: Evaluating Large Language Models Toward Improved Medical Concept Extraction

Medical concept extraction from electronic health records underpins many downstream applications, yet remains challenging because medically meaningful concepts are frequently implied rather than explicitly stated in medical narratives. Existing benchmarks with human-annotated evidence spans underscore the importance of grounding extracted concepts in medical text. However, they predominantly focus on explicitly stated concepts instead of implicit concepts. We present MedicalBench, a benchmark for medical concept extraction with evidence grounding that evaluates implicit medical reasoning. MedicalBench formulates medical concept extraction as a verification task over medical note-concept pairs, coupled with sentence-level evidence identification. Built from MIMIC-IV discharge summaries and human-verified ICD-10 codes, the dataset is curated through a multi-stage large language model (LLM) triage pipeline followed by medical annotation and expert review. It deliberately includes implicit positives, semantically confusable negatives, and cases where LLM judgments disagree with medical expert assessments. We define two complementary evaluation tasks: (1) medical concept extraction and (2) sentence-level evidence retrieval, enabling assessment of both correctness and interpretability. Benchmarking state-of-the-art LLMs reveals that performance remains modest, highlighting the difficulty of extracting implicitly expressed concepts. We further show that performance is largely invariant to note length, indicating that MedicalBench isolates reasoning difficulty rather than superficial confounders. MedicalBench provides the first systematic benchmark for implicit, evidence-grounded medical concept extraction, offering a foundation for developing medical language models that can both identify medically relevant concepts and justify their predictions in a transparent and medically faithful manner.

cs.CL↗

arXiv2026

Fine-grained Image Quality Assessment for Perceptual Image Restoration

Recent years have witnessed remarkable achievements in perceptual image restoration (IR), creating an urgent demand for accurate image quality assessment (IQA), which is essential for both performance comparison and algorithm optimization. Unfortunately, the existing IQA metrics exhibit inherent weakness for IR task, particularly when distinguishing fine-grained quality differences among restored images. To address this dilemma, we contribute the first-of-its-kind fine-grained image quality assessment dataset for image restoration, termed FGRestore, comprising 18,408 restored images across six common IR tasks. Beyond conventional scalar quality scores, FGRestore was also annotated with 30,886 fine-grained pairwise preferences. Based on FGRestore, a comprehensive benchmark was conducted on the existing IQA metrics, which reveal significant inconsistencies between score-based IQA evaluations and the fine-grained restoration quality. Motivated by these findings, we further propose FGResQ, a new IQA model specifically designed for image restoration, which features both coarse-grained score regression and fine-grained quality ranking. Extensive experiments and comparisons demonstrate that FGResQ significantly outperforms state-of-the-art IQA metrics. Codes and model weights have been released in https://sxfly99.github.io/FGResQ-Home.

eess.IV↗

arXiv2026

Fine-grained Image Aesthetic Assessment: Learning Discriminative Scores from Relative Ranks

Image aesthetic assessment (IAA) has extensive applications in content creation, album management, and recommendation systems, etc. In such applications, it is commonly needed to pick out the most aesthetically pleasing image from a series of images with subtle aesthetic variations, a topic we refer to as fine-grained IAA. Unfortunately, state-of-the-art IAA models are typically designed for coarse-grained evaluation, where images with notable aesthetic differences are evaluated independently on an absolute scale. These models are inherently limited in discriminating fine-grained aesthetic differences. To address the dilemma, we contribute FGAesthetics, a fine-grained IAA database with 32,217 images organized into 10,028 series, which are sourced from diverse categories including Natural, AIGC, and Cropping. Annotations are collected via pairwise comparisons within each series. We also devise Series Refinement and Rank Calibration to ensure the reliability of data and labels. Based on FGAesthetics, we further propose FGAesQ, a novel IAA framework that learns discriminative aesthetic scores from relative ranks through Difference-preserved Tokenization (DiffToken), Comparative Text-assisted Alignment (CTAlign), and Rank-aware Regression (RankReg). FGAesQ enables accurate aesthetic assessment in fine-grained scenarios while still maintains competitive performance in coarse-grained evaluation. Extensive experiments and comparisons demonstrate the superiority of the proposed method.

cs.CV↗

arXiv2026

TARSE: Test-Time Adaptation via Retrieval of Skills and Experience for Reasoning Agents

Complex clinical decision making often fails not because a model lacks facts, but because it cannot reliably select and apply the right procedural knowledge and the right prior example at the right reasoning step. We frame clinical question answering as an agent problem with two explicit, retrievable resources: skills, reusable clinical procedures such as guidelines, protocols, and pharmacologic mechanisms; and experience, verified reasoning trajectories from previously solved cases (e.g., chain-of-thought solutions and their step-level decompositions). At test time, the agent retrieves both relevant skills and experiences from curated libraries and performs lightweight test-time adaptation to align the language model's intermediate reasoning with clinically valid logic. Concretely, we build (i) a skills library from guideline-style documents organized as executable decision rules, (ii) an experience library of exemplar clinical reasoning chains indexed by step-level transitions, and (iii) a step-aware retriever that selects the most useful skill and experience items for the current case. We then adapt the model on the retrieved items to reduce instance-step misalignment and to prevent reasoning from drifting toward unsupported shortcuts. Experiments on medical question-answering benchmarks show consistent gains over strong medical RAG baselines and prompting-only reasoning methods. Our results suggest that explicitly separating and retrieving clinical skills and experience, and then aligning the model at test time, is a practical approach to more reliable medical agents.

cs.IR↗

arXiv2026

Fast and Effective On-policy Distillation from Reasoning Prefixes

On-policy distillation (OPD), which samples trajectories from the student model and supervises them with a teacher at the token level, avoids relying solely on verifiable terminal rewards and can yield better generalization than off-policy distillation. However, OPD requires expensive on-the-fly sampling of the student policy during training, which substantially increases training cost, especially for long responses. Our initial analysis shows that, during OPD, training signals are often concentrated in the prefix of each output, and that even a short teacher-generated prefix can significantly help the student produce the correct answer. Motivated by these observations, we propose a simple yet effective modification of OPD: we apply the distillation objective only to prefixes of student-generated outputs and terminate each sampling early during distillation. Experiments on a suite of AI-for-Math and out-of-domain benchmarks show that on-policy prefix distillation matches the performance of full OPD while reducing training FLOP by 2x-47x.

cs.LG↗